Study On Terrain Recognition And Obstacle Avoidance For Off-road Mobile Robot

For mobile robot, it is the sensing ability of environment that largely decides its ability of autonomous moving in environment and affects its overall performance assessment. And meanwhile, the terrain classification and how to deal with it must be solved for mobile robot. Outdoor mobile robot moves mainly in urban outdoor structured environment and wild unstructured environment, and thus to identify and classify complex terrain, to analyze the centimeter-level obstacle qualitatively and make the appropriate avoidance principle and path planning had been one of most important technical issues in autonomous mobile robots.In this paper, using outdoor mobile robot with Kinect as research platform, we propose obstacle avoidance principle and do some work in outdoor terrain recognition and classification in the absence of sunlight.As terrain features are always complex and difficult to identify using one single feature, we combine many features, such as waviness, gradient, texture, color, together to solve the recognition and classification problems. In this paper, we divide terrain into four types:artificial rigid pavement, natural road, vegetation and water.When environment is complex, the2D grid map cannot present so much information and the3D grid map has relatively high computational complexity. So we can use the2.5D grid map to represent much information such as elevation, terrain features, the position and size of obstacle in the scene. With these information, we can compute the traversability region and make appropriate obstacle avoidance strategy。 Then we use training methods to determine the traversability of kinds of terrains.High computational complexity and poor robustness in accessing depth information are two main problems for stereo vision. In this paper, we use kinect as a sensor, which uses coded structured light method to get the depth of information and uses the depth information to access the size of obstacles within a scene.Finally, the problems encountered during the completion of the paper are summarized, and according to these, thinking of the problems and directions helping solve these problems is as well as proposed.